P7UC2024 - Multidimensional optimization of loudspeakers

Download insert_drive_file

Authors:

Alfred J. Svobodnik, CEO, Mvoid Audio Technologies

Journal / Event:

pSeven User Conference 2024

Abstract:

By combining a parametric multiphysics model with an optimization algorithm, the performance of electrodynamic converters is being improved. To do this, the optimization strategy varies the geometric parameters of the loudspeaker, whose target functions are recalculated in the multiphysics FEA model. The optimizer processes each FEA solution of a variation and thus builds a surrogate model, which is successively trained by comparison. The optimizer controls the parameter variations for fast convergence. The result is a significant improvement in the nonlinear parameters that affect sound quality, i.e. the nonlinear distortion, that characterizes the large-signal behavior of the speaker.

Keywords: P7UC2024, Electronics, Machine Learning, Optimization